What data an app collects, and why

You hire a bike with an app on your phone. Before the bike unlocks, the app has your name, your email and a bank card.

When you hand the bike back, it knows where you rode and for how long. Why does it need all of that? And who else gets to see it?

A service should collect data only for a purpose it can name. The bike app needs your card to charge you. It needs the bike's place to find it again.

Does it need your date of birth? Perhaps, to check you are old enough to ride. Does it need your phone's contacts? No. Collecting only what a purpose needs is called data minimisation.

docker exec jupyter head -4 hires.csv

The first line names the columns: the date, the month, the weekday, the weather, the temperature, the rain and the hires. Each line after it is one day.

Nothing says who rode. The shop wants to know how many bikes to put out each morning. A count for each day answers that, so it never needs a name.

docker exec jupyter python3 steps/load.py

Personal data is any data that can be linked to one person. A name is personal data. So is an email address, a face in a photo and a phone's location.

Some data only points to a person when it is put together. A postcode alone covers many homes. A postcode, a birth date and the time somebody always rides can point to one person.

In the UK and the EU, the law says a service needs a fair reason to use your personal data. Consent is one reason. A contract is another, such as an address for a delivery.

The service must tell you what it collects and why, in plain words. It must keep the data safe, and only for as long as it needs it. You can ask to see what it holds about you.

To study data safely, anonymise it. Delete names and emails. Make exact values rougher: an age of 17 becomes the band 10 to 19, and a postcode keeps only its first part.

Swapping names for codes is not enough. That is called pseudonymised data, because whoever keeps the list of codes can put the names back.

# One rider's record, as the app stored it.
rider = {"name": "Priya Shah", "postcode": "LS6 2AB", "age": 17, "rides": 12}

# Keep only what the study needs, and make it rougher.
low = rider["age"] // 10 * 10
safe = {}
safe["area"] = rider["postcode"].split(" ")[0]
safe["age_band"] = str(low) + "-" + str(low + 9)
safe["rides"] = rider["rides"]
print(safe)

# Change the age to 34 and the postcode to "M1 4BT".
# Press Run again. What changes in the answer?

import pandas as pd

riders = pd.DataFrame({"name": ["Priya", "Tom", "Ana"],
                       "postcode": ["LS6 2AB", "LS1 4DY", "M1 4BT"],
                       "age": [17, 34, 17],
                       "rides": [12, 3, 8]})
# Keep the first part of each postcode, then drop what names a person.
riders["area"] = riders["postcode"].str.split(" ").str[0]
riders.drop(columns=["name", "postcode"])

Open JupyterLab from the bar, make a new notebook and type the code above into a cell. Press Shift and Enter.

Then pick an app you use. Find its privacy notice and list three pieces of data it collects. Next to each, write the purpose it gives. Is any item one it does not need?